Page 305 - Numerical Analysis and Modelling in Geomechanics

286 ANALYSIS AND DESIGN OF PILE GROUPS
until the load corresponding to the yield of the first interface element is reached;
(2) deformations are often seriously underestimated at high load levels. An
alternative approach is offered by the widely used computer program MPILE,
originally developed by Randolph (1980) under the name of PIGLET. The
analysis is based on a semi-empirical method which makes use of approximate
analytical solutions for single pile response and for interaction between two piles,
in which linear elastic soil behaviour is assumed.
It is important to note that the interaction factor approach (such as is employed
in DEFPIG and MPILE) solves the group problem by calculating the influence
coefficients for each pair of piles and by merely superimposing the effects. This
approximate procedure produces a number of limitations: (a) it ignores the
stiffening effect of intervening piles in a group, thereby leading to an
overestimation of interaction between piles; (b) its use becomes questionable for
cases in which not all the piles are identical; (c) it only gives the loads and
bending moments at the pile heads, but not their distributions along the piles;
these may only be approximated utilising the single pile solutions with the
corresponding pile head loads and bending moments.
The above limitations on the use of interaction factors may be removed by
simultaneous consideration of all the piles within the group, i.e. performing a
“complete” analysis of the group. The computer program PGROUP, originally
developed by Banerjee and Driscoll (1976), is included in this category but is
restricted to linear elastic analyses and problems of small dimensions because of
the very large computational resources required. The latter aspect makes the
program inapplicable in normal design. An even more rigorous linear analysis is
performed by the numerical code GEPAN (Xu and Poulos, 2000) in which the
boundary elements are meshed in partly cylindrical or annular surfaces. The
program provides a benchmark for assessing the accuracy of simplified
procedures in the linear range and can also analyse loadings induced by ground
movements. However, the relatively high computational cost makes questionable
its potential use for routine design problems.
The main feature of the proposed PGROUPN program (Basile, 1999) lies in
its capability to provide a complete non-linear BEM solution of the soil
continuum while retaining a computationally efficient code. One of the main
advantages of a non-linear analysis system over a linear one is that it has the
desirable effect of demonstrating a relative reduction of the corner loads in pile
groups in both the vertical and horizontal senses. This observation is of basic
importance in practice, and offers the prospect of significant improvements in
design techniques and potential saving of materials. The choice of soil
parameters for PGROUPN is simple and direct: for a linear analysis, it is only
necessary to define two soil parameters whose physical interpretation is clear,
i.e. the soil modulus (E ) and the Poisson’s ratio (v ). If the effects of soil non-
s
s
linearity are considered, the strength properties of the soil also need to be
specified, i.e. the undrained shear strength (C ) for cohesive soils and the angle
u
of friction (ф') for cohesionless soils. These parameters are routinely measured in